1. Executive Summary

This capstone project, part of the Google Data Analytics Certificate, analyzes the fictional company Cyclistics 2024 bike-share data to understand how casual riders and annual members use the service differently, informing a marketing strategy focused on conversion. Key findings highlight distinct usage patterns to guide targeted efforts.

Trip Patterns: Casual riders take longer, weekend-heavy recreational trips; members have shorter, consistent commutes. Seasonality: Casual rides peak in July–August, dropping sharply post-summer; members maintain steadier year-round engagement. Timing: Members peak during commute hours (7–9 AM, 4–6 PM); casual riders favor midday (11 AM–3 PM) and weekends. Stations: Casual riders frequent tourist areas (e.g., Streeter Dr & Grand Ave); members use business hubs (e.g., Kingsbury St & Kinzie St). Ride Type: Classic bikes dominate for both groups; casual riders show greater use of electric bikes and scooters. Recommendations: Use a location service API (Google used in in this analysis) to target marketing at tourist-heavy stations during summer, promoting annual memberships to casual riders by emphasizing year-round utility.

2. About Cyclistic

Cyclistic operates a fleet of 5,824 bicycles across 692 docking stations in Chicago. The company offers flexible pricing options including single rides, day passes, and annual memberships.

Approximately 30% of users commute to work.
The remainder primarily use bikes for leisure.

3. Key Roles

Lily Moreno – Director of Marketing
Marketing Analytics Team – Internal analysts studying user behavior
Executive Team – Final decision makers on marketing initiatives

4. Business Task

The marketing team has posed three strategic questions:

How do annual members and casual riders use Cyclistic bikes differently?
Why would casual riders consider becoming annual members?
How can digital media influence casual riders to convert?

This project focuses on answering Question 1 through deep analysis of ride data.

Specific Questions Answered

What is the average ride duration for each rider type?
How does ride duration vary by day of week and by month?
What is the total number of rides per day for both groups?
What are the busiest ride hours?
Do patterns differ between weekdays and weekends?
What type of bike is most used by each group?
What stations are most used overall and by user type?

5. Data Source, Tools and Setup

This section documents the dataset origin, tools used, and the creation of a working dataset to support all downstream validation and analysis.

Show Details

5.1 Data Source and Credibility

Cyclistics data was downloaded from the Divvy Bikes public portal. It covers January through December 2024 and includes a total of 5,860,568 ride records. The dataset is assumed complete and reliable for analysis.

5.2 Tools and Packages

All analysis was conducted in R using RStudio. The following packages were used:

Data Cleaning and Transformation
- dplyr, tidyr – data manipulation
- janitor – column name cleanup and frequency tables
- lubridate – datetime parsing

Visualization
- ggplot2 – charting
- scales – axis label formatting

Reporting and Output
- knitr, kableExtra – table and report rendering
- skimr – compact column summaries

Show Package Installation Commands

# install.packages("tidyverse")
# install.packages("readr")
# install.packages("lubridate")
# install.packages("dplyr")
# install.packages("janitor")
# install.packages("knitr")
# install.packages("tidyr")
# install.packages("scales")
# install.packages("kableExtra")
# install.packages("skimr")

5.3 Load Raw Dataset

The raw data is loaded into original_data, then previewed to verify structure and dimensions.

Show Code

# Load raw, combined dataset from CSV file
original_data <- read_csv("C:\\Users\\jonwd\\Documents\\R_docs\\Cyclistics_Project_Data\\Capstone_files\\original_data.csv")

# Track row count
row_count_original <- nrow(original_data)

# Preview first 5 rows instead of glimpse
head(original_data, 5)

## # A tibble: 5 × 13
##   ride_id          rideable_type started_at          ended_at           
##   <chr>            <chr>         <dttm>              <dttm>             
## 1 C1D650626C8C899A electric_bike 2024-01-12 15:30:27 2024-01-12 15:37:59
## 2 EECD38BDB25BFCB0 electric_bike 2024-01-08 15:45:46 2024-01-08 15:52:59
## 3 F4A9CE78061F17F7 electric_bike 2024-01-27 12:27:19 2024-01-27 12:35:19
## 4 0A0D9E15EE50B171 classic_bike  2024-01-29 16:26:17 2024-01-29 16:56:06
## 5 33FFC9805E3EFF9A classic_bike  2024-01-31 05:43:23 2024-01-31 06:09:35
## # ℹ 9 more variables: start_station_name <chr>, start_station_id <chr>,
## #   end_station_name <chr>, end_station_id <chr>, start_lat <dbl>,
## #   start_lng <dbl>, end_lat <dbl>, end_lng <dbl>, member_casual <chr>

Results: - The raw dataset was successfully loaded into ‘original_data’ with a total of 5,860,568 rows.

5.4 Create Working Dataset

To support consistent filtering and analysis, we create a new working dataset (‘cyclistics_data’) that includes the following derived fields:

ride_length: Ride duration in whole minutes (rounded)
day_of_week: Full weekday name (week starts on Sunday)
hour_of_day: Ride start hour (0–23)
month: Labeled month (e.g., January)
These fields are computed from the original timestamp columns.

Show Code

cyclistics_data <- original_data %>%
  mutate(
    ride_length = as.numeric(difftime(ended_at, started_at, 
                                      units = "mins")),
    ride_length = round(ride_length, 0),
    day_of_week = wday(started_at, label = TRUE, abbr = FALSE, 
                        week_start = 7),
    hour_of_day = hour(started_at),
    month = month(started_at, label = TRUE, abbr = FALSE)
  )

Results: The new working dataset, ‘cyclistics_data’, includes all original fields plus the four derived columns listed above. This dataset will be used for all further validation and analysis. Row count: 5,860,568

6. Data Cleaning

This section ensures the dataset is reliable for analysis by applying validation and preparation steps to remove incomplete, invalid, or untrustworthy records. Each step addresses specific data quality issues to support accurate insights into Cyclistic’s rider patterns.

Show Details

6.1 Identify Rows with Insufficient Data

We check for empty strings in critical fields (ride_id, started_at, ended_at, member_casual, rideable_type) in cyclistics_data, replacing them with NA for consistency. Rows missing essential fields (ride_id, started_at, ended_at, member_casual) are identified for later removal.

Show Code

# Define critical fields
critical_fields <- c("ride_id", "started_at", "ended_at", "member_casual", 
                     "rideable_type")

# Note: cyclistics_data is defined in Section 5.4
# Identify character columns among critical fields
char_fields <- critical_fields[sapply(cyclistics_data[critical_fields], is.character)]

# Check for empty strings in character fields
if (length(char_fields) == 0) {
  has_empty <- logical(0)
  message("No character fields to check for empty strings; skipping transformation.")
} else {
  has_empty <- sapply(char_fields, function(col) {
    any(cyclistics_data[[col]] == "", na.rm = TRUE)
  })
  if (any(has_empty)) {
    fields_to_transform <- char_fields[has_empty]
    cyclistics_data <- cyclistics_data %>%
      mutate(
        across(
          all_of(fields_to_transform),
          ~ na_if(trimws(.), "")
        )
      )
  } else {
    message("No empty strings found in character fields; skipping transformation.")
  }
}

# Identify rows with insufficient data (missing critical fields)
insufficient_rows <- cyclistics_data %>%
  filter(
    is.na(ride_id) | 
    is.na(started_at) | 
    is.na(ended_at) | 
    is.na(member_casual)
  ) %>%
  nrow()

# Free memory
gc()

##             used  (Mb) gc trigger   (Mb)  max used   (Mb)
## Ncells   7061623 377.2   12727297  679.8   7552518  403.4
## Vcells 120523668 919.6  446364176 3405.5 490626680 3743.2

Results:

Empty strings in critical fields (ride_id, started_at, ended_at, member_casual, rideable_type) were replaced with NA. Identified 44 rows with insufficient data (missing ride_id, started_at, ended_at, or member_casual) for later removal. Row count unchanged: 5,860,568.

6.2 Verify Data Types

We verify that the started_at and ended_at fields in cyclistics_data are properly formatted datetime objects, enabling accurate time-based calculations (e.g., ride durations).

Show Code

library(tidyverse) # Load tidyverse for %>%
# Check the class of started_at and ended_at
data.frame(
  Column = c("started_at", "ended_at"),
  Class = c(class(cyclistics_data$started_at)[1], class(cyclistics_data$ended_at)[1])
) %>%
  knitr::kable() %>%
  kableExtra::kable_styling(full_width = FALSE, position = "left")

Column	Class
started_at	POSIXct
ended_at	POSIXct

Results:
Confirmed started_at and ended_at are POSIXct, a format R uses to store dates and times for accurate calculations, like ride durations and monthly patterns. Row count unchanged: 5,860,568

6.3 Verify Categorical Fields

We verify that rideable_type and member_casual fields in cyclistics_data are standardized to ensure consistent visualizations and reporting.

Show Code

# Compute and display unique levels
rideable_levels <- unique(cyclistics_data$rideable_type)
member_levels <- unique(cyclistics_data$member_casual)

cat("Unique rideable_type values:", paste(rideable_levels, collapse = ", "), "\n")

## Unique rideable_type values: electric_bike, classic_bike, electric_scooter

cat("Unique member_casual values:", paste(member_levels, collapse = ", "), "\n")

## Unique member_casual values: member, casual

# Free memory
gc()

##             used  (Mb) gc trigger   (Mb)  max used   (Mb)
## Ncells   6952190 371.3   10398622  555.4   7060364  377.1
## Vcells 120305764 917.9  402365447 3069.9 345861957 2638.8

Results:
Confirmed rideable_type values: electric_bike, classic_bike, electric_scooter. Confirmed member_casual values: member, casual. Row count unchanged: 5,860,568.

6.4 Validate Minimum Required Fields

We check for missing values in key fields of cyclistics_data required for ride-level analysis or mapping.

Show Code

# Summarize missing values
missing_data <- cyclistics_data %>%
  summarise(
    `Ride ID` = sum(is.na(ride_id)),
    `Rideable Type` = sum(is.na(rideable_type)),
    `Start Station` = sum(is.na(start_station_name) & is.na(start_station_id)),
    `End Station` = sum(is.na(end_station_name) & is.na(end_station_id)),
    `Start Time` = sum(is.na(started_at)),
    `End Time` = sum(is.na(ended_at)),
    `User Type` = sum(is.na(member_casual)),
    `Start Coordinates` = sum(is.na(start_lat) & is.na(start_lng)),
    `End Coordinates` = sum(is.na(end_lat) & is.na(end_lng))
  )

# Convert to long format and filter non-zero missing values
missing_data_long <- missing_data %>%
  pivot_longer(everything(), names_to = "Field", values_to = "Missing Count") %>%
  filter(`Missing Count` > 0)

# Display as a table (if there are missing values)
if (nrow(missing_data_long) > 0) {
  missing_data_long %>%
    knitr::kable() %>%
    kableExtra::kable_styling(full_width = FALSE, position = "left")
} else {
  cat("No missing values found in required fields.\n")
}

Field	Missing Count
Start Station	1073951
End Station	1104653
End Coordinates	7232

Results:
Fields with missing values: Start Station: 1,073,951 rows End Station: 1,104,653 rows Start Time and End Time: 44 rows each End Coordinates: 7,232 rows Other fields (Ride ID, Rideable Type, User Type, Start Coordinates) have no missing values. Row count unchanged: 5,860,568.

6.5 Count Records to be Removed

We calculate the number and proportion of rows that fail any critical data checks.

Show Code

rows_to_remove <- cyclistics_data %>%
  filter(
    is.na(ride_id) |
    is.na(started_at) |
    is.na(ended_at) |
    is.na(member_casual) |
    (is.na(start_station_name) & is.na(start_station_id)) |
    (is.na(end_station_name) & is.na(end_station_id))
  ) %>%
  summarise(rides_to_remove = n()) %>%
  pull(rides_to_remove)

total_rows <- nrow(cyclistics_data)
removal_ratio <- rows_to_remove / total_rows

data.frame(
  `Rows to Remove` = format(rows_to_remove, big.mark = ","),
  `Total Rows` = format(total_rows, big.mark = ","),
  `Removal Ratio` = scales::percent(removal_ratio, accuracy = 0.01)
) %>%
  knitr::kable() %>%
  kableExtra::kable_styling(full_width = FALSE)

Rows.to.Remove	Total.Rows	Removal.Ratio
1,652,259	5,860,568	28.19%

Results: Rows to remove: 1,652,289 Total rows: 5,860,568 Removal ratio: 28.20% These records, often missing station identifiers or timestamps, will be excluded to ensure reliable analysis.

6.6 Check for Duplicate Ride IDs

We check for duplicate ride_id values in cyclistics_data, as each ride should have a unique identifier.

Show Code

# Count occurrences of each ride_id and identify duplicates
duplicate_stats <- cyclistics_data %>%
  count(ride_id) %>%
  filter(n > 1)

# Number of unique ride_ids with duplicates
duplicate_ids <- nrow(duplicate_stats)

# Total number of duplicate records (beyond the first occurrence)
duplicate_records <- sum(duplicate_stats$n) - nrow(duplicate_stats)

cat(format(duplicate_ids, big.mark = ","), "unique `ride_id`s with duplicates identified, totaling", 
    format(duplicate_records, big.mark = ","), "duplicate records.\n")

## 211 unique `ride_id`s with duplicates identified, totaling 211 duplicate records.

Results:
Identified 211 unique ride_ids with duplicates, totaling 211 duplicate records. These will be removed (keeping the first occurrence) to maintain data integrity. Row count unchanged: 5,860,568.

6.7 Identify Short Rides

We identify rides in cyclistics_data with a duration of 1 minute or less, which may indicate system tests or early cancellations.

Show Code

short_rides <- cyclistics_data %>%
  filter(ride_length <= 1) %>%
  summarise(short_rides = n()) %>%
  pull(short_rides)

cat(format(short_rides, big.mark = ","), "rides identified.\n")

## 173,661 rides identified.

Results:
Identified 173,661 rides with a duration of 1 minute or less, likely system tests or cancellations. These will be removed to ensure data quality. Row count unchanged: 5,860,568.

6.8 Apply Validation Filters and generate data for exploratory analysis.

We now apply all previously defined filters to remove incomplete, duplicate, or low-quality records.

Show Code

eda_data <- cyclistics_data %>%
  # Remove duplicates
  distinct(ride_id, .keep_all = TRUE) %>%
  # Apply filters for missing critical fields
  filter(
    !is.na(ride_id),
    !is.na(started_at),
    !is.na(ended_at),
    !is.na(member_casual),
    !(is.na(start_station_name) & is.na(start_station_id)),
    !(is.na(end_station_name) & is.na(end_station_id)),
    # Exclude short rides and negative durations
    ride_length > 1,
    ride_length >= 0
  )

rows_removed <- nrow(cyclistics_data) - nrow(eda_data)
removal_ratio <- rows_removed / nrow(cyclistics_data)

data.frame(
  `Final Row Count` = format(nrow(eda_data), big.mark = ","),
  `Rows Removed` = format(rows_removed, big.mark = ","),
  `Removal Ratio` = scales::percent(removal_ratio, accuracy = 1)
) %>%
  knitr::kable() %>%
  kableExtra::kable_styling(full_width = FALSE)

Final.Row.Count	Rows.Removed	Removal.Ratio
4,142,075	1,718,493	29%

# Save the cleaned dataset to a new CSV file
write_csv(eda_data, "C:\\Users\\jonwd\\Documents\\R_docs\\Cyclistics_Project_Data\\Capstone_files\\eda_data.csv")

Results:
Final dataset (eda_data): 4,142,075 rows. Removed: 1,718,493 rows (29.33% of original dataset). Many rows failed multiple criteria (e.g., missing stations, duplicates, short rides), ensuring a robust dataset for analysis.

7. Analysis and Findings

This section analyzes eda_data to compare casual riders and annual members’ usage patterns, addressing Section 4’s business questions. We trim rides of extreme duration, then explore ride length, daily/monthly patterns, peak hours, weekday/weekend differences, bike preferences, and popular stations using summary statistics and ggplot2 visualizations.

Show Details

7.1 Initial Look - Outliers

We use the Interquartile Range (IQR) to identify unusually short or long ride durations, calculated as rides below Q1 - 1.5 * IQR or above Q3 + 1.5 * IQR. The IQR method spots unusual data by measuring the spread of the middle 50% of ride times. It flags rides far beyond this range as outliers, helping us focus on common patterns without complex assumptions.

Show Code

# Ensure ride_length is numeric
eda_data <- eda_data %>% mutate(ride_length = as.numeric(ride_length))

# Calculate IQR bounds
q1 <- quantile(eda_data$ride_length, 0.25)
q3 <- quantile(eda_data$ride_length, 0.75)
iqr <- q3 - q1
lower_bound <- max(0, q1 - 1.5 * iqr)  # Prevent negative durations
upper_bound <- q3 + 1.5 * iqr

# Permanently add is_outlier column to eda_data
eda_data <- eda_data %>%
  mutate(is_outlier = ride_length > upper_bound)

# Count outliers
outliers <- eda_data %>% filter(ride_length < lower_bound | ride_length > upper_bound)
outlier_count <- nrow(outliers)
total_rides <- nrow(eda_data)

# Display results
cat("Q1 (25th percentile):", round(q1, 1), "minutes\n")

## Q1 (25th percentile): 6 minutes

cat("Q3 (75th percentile):", round(q3, 1), "minutes\n")

## Q3 (75th percentile): 18 minutes

cat("IQR:", round(iqr, 1), "minutes\n")

## IQR: 12 minutes

cat("Lower bound:", round(lower_bound, 1), "minutes\n")

## Lower bound: 0 minutes

cat("Upper bound:", round(upper_bound, 1), "minutes\n")

## Upper bound: 36 minutes

cat("Total rides:", format(total_rides, big.mark = ","), "\n")

## Total rides: 4,142,075

cat("Outlier rides:", format(outlier_count, big.mark = ","), "\n")

## Outlier rides: 333,428

cat("Outlier percentage:", scales::percent(outlier_count / total_rides, accuracy = 0.1), "\n")

## Outlier percentage: 8.0%

Results:
The IQR method flags 333,428 rides (8.0% of 4,142,075) with durations ≥36 minutes as outliers. Section 6.8 removed all rides under 2 minutes in duration. The bar chart below categorizes ride durations, distinguishing typical rides (≤36 minutes) from outliers (>36 minutes), ensuring subsequent analyses focus on common rider behavior.

Show Code

library(ggplot2)
library(dplyr)
library(ggbreak)
library(scales)

total_rides <- nrow(eda_data)
ride_bins <- eda_data %>%
  mutate(
    bin = case_when(
      ride_length <= 36.0 ~ "≤ 36.0 minutes (Retained)",
      ride_length > 36.0 ~ "> 36.0 minutes (Outliers)"
    ),
    bin = factor(bin, levels = c("≤ 36.0 minutes (Retained)", "> 36.0 minutes (Outliers)"))
  ) %>%
  group_by(bin) %>%
  summarise(count = n()) %>%
  mutate(
    percent = count / total_rides * 100,
    label = sprintf("%s (%s%%)", comma(count), round(percent, 1)),
    vjust = -0.8
  )

# Store plot
duration_bar_plot <- ggplot(ride_bins, aes(x = bin, y = count, fill = bin)) +
  geom_bar(stat = "identity", color = "white") +
  geom_text(aes(label = label, vjust = vjust), size = 4) +
  scale_fill_manual(values = c(
    "≤ 36.0 minutes (Retained)" = "darkorange",
    "> 36.0 minutes (Outliers)" = "grey"
  )) +
  labs(
    title = "Ride Duration Categories (IQR Method)",
    x = "Ride Duration Category",
    y = "Number of Rides",
    fill = "Category",
    caption = "> 36.0 minutes truncated; y-axis compressed."
  ) +
  scale_y_continuous(limits = c(0, 4800000), labels = comma, expand = expansion(mult = c(0, 0.1))) +
  scale_y_break(c(400000, 3000000), scales = "free", space = 0.2) + # to truncate chart for clarity
  theme_minimal(base_size = 11) +
  theme(
    axis.text.y = element_text(size = 10),
    legend.position = "bottom",
    axis.title.x = element_text(hjust = 0.5),
    plot.caption = element_text(hjust = 0.5)
  )

# Tag outliers within eda_data
eda_data <- eda_data %>%
  mutate(is_outlier = ride_length > 36.0)

# Count flagged and retained rides
outlier_total <- sum(eda_data$is_outlier)
valid_total <- sum(!eda_data$is_outlier)
total_rows <- nrow(eda_data)
outlier_pct <- round(100 * outlier_total / total_rows, 1)

Results: All subsequent analyses will exclude 333,428 outlier rides (≥36 minutes) using filter(!is_outlier), focusing on typical patterns to inform marketing strategies. Long rides warrant future investigation to understand niche usage behaviors.

7.2 Monthly Ride Volume by rider type

This analysis excludes outlier rides longer than 36 minutes (as flagged in Section 7.2). The line chart below highlights the top 3 months for each rider group using circular markers and floating labels. This style echoes the visual language used in the hourly dot chart later in the report.

Show Code

library(dplyr)
library(ggplot2)
library(scales)
library(lubridate)

# Define colors locally
cyclistics_colors <- c(
  "member" = "#1f77b4",   # Blue
  "casual" = "#ff7f0e"    # Orange
)

# Prepare data: filter outliers, summarize by month and rider type
monthly_data <- eda_data %>%
  filter(!is_outlier) %>%
  mutate(month = month(started_at, label = TRUE, abbr = FALSE)) %>%
  group_by(member_casual, month) %>%
  summarise(rides = n(), .groups = "drop") %>%
  mutate(
    month = factor(month, levels = month.name),
    member_casual = factor(member_casual, levels = c("casual", "member")),
    
  )

# Identify top 3 months for each rider type
monthly_labels <- monthly_data %>%
  group_by(member_casual) %>%
  slice_max(order_by = rides, n = 3) %>%
  ungroup() %>%
  mutate(label = comma(rides))

# Build the plot
monthly_line_plot <- ggplot(monthly_data, aes(x = month, y = rides, group = member_casual, color = member_casual)) +
  geom_line(size = 1.2) +
  geom_point(
    data = monthly_labels,
    aes(x = month, y = rides),
    shape = 21,
    fill = "white",
    stroke = 1.2,
    size = 4,
    inherit.aes = FALSE
  ) +
  geom_label(
    data = monthly_labels,
    aes(x = month, y = rides + 20000, label = label),
    fill = "white",
    label.size = 0.2,
    size = 3.5,
    fontface = "bold",
    label.padding = unit(0.2, "lines"),
    label.r = unit(0.2, "lines"),
    inherit.aes = FALSE,
    show.legend = FALSE
  ) +
  scale_color_manual(values = cyclistics_colors) +
  scale_y_continuous(labels = comma, limits = c(0, 340000)) +
  labs(
    title = "Monthly Ride Volume by Rider Type",
    subtitle = "Top 3 months labeled",
    x = "Month",
    y = "Total Rides",
    color = "Rider Type"
  ) +
  theme_minimal(base_size = 11) +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1),
    plot.title = element_text(face = "bold", hjust = 0.5),
    plot.subtitle = element_text(hjust = 0.5),
    legend.position = "bottom"
  )

# Save plot
ggsave("monthly_lineplot_top3_v9.png", plot = monthly_line_plot, width = 8, height = 4.8, dpi = 150)

Results:
Both rider groups show strong seasonal trends, with casual usage peaking in July and August and member usage peaking slightly later in September. Casual rides drop sharply after summer, while member rides remain higher into the fall, indicating more consistent year-round engagement. These trends suggest casual riders are primarily recreational, while members include commuters or frequent users less sensitive to seasonality

7.3 Total Rides by Rider Type

To compare overall usage by rider type, we calculated total rides for casual and member riders, excluding rides longer than 36 minutes, as flagged in Section 7.2. The donut chart and summary table below reflect only rides considered typical (ride_length ≤ 36.0).

Show Code

library(ggplot2)
library(dplyr)
library(scales)

output_dir <- "Cyclistics_Project_Data/V9"
if (!dir.exists(output_dir)) dir.create(output_dir, recursive = TRUE)
output_file <- file.path(output_dir, "rider_type_summary.csv")

rider_type_summary <- eda_data %>%
  filter(!is_outlier) %>%  # <-- new line: exclude outlier rides
  mutate(member_casual = tolower(trimws(member_casual))) %>%   # <--- NEW: enforce color match
  group_by(member_casual) %>%
  summarise(total_rides = n(), .groups = "drop") %>%
  mutate(
    proportion = total_rides / sum(total_rides),
    percent_label = sprintf("%s\n%s\n%s%%", member_casual, format(total_rides, big.mark = ","), round(proportion * 100)),
    total_rides_label = format(sum(total_rides), big.mark = ",")
  )

rider_type_summary %>%
  select(member_casual, total_rides, percent = proportion) %>%
  mutate(
    total_rides = format(total_rides, big.mark = ","),
    percent = scales::percent(percent, accuracy = 1)
  ) %>%
  knitr::kable(caption = "Total Rides by Rider Type") %>%
  kableExtra::kable_styling(full_width = FALSE, position = "left")

Total Rides by Rider Type
member_casual	total_rides	percent
casual	1,261,117	33%
member	2,547,530	67%

write_csv(rider_type_summary, output_file)

Results:
The table summarizes total rides and percentages for each rider type. The dataset rider_type_summary.csv is exported for Tableau. The donut chart visualizes the proportion of rides, with members typically having a higher share (e.g., ~60%) than casual riders (e.g., ~40%).

7.4 Average Ride Duration by Day

This section examines the average ride duration for casual riders and annual members by day of the week, focusing on typical rides (ride_length <= 36.0 minutes) to exclude outliers.

Show Code

library(ggplot2)
library(dplyr)
library(scales)

output_dir <- "Cyclistics_Project_Data/V9"
if (!dir.exists(output_dir)) dir.create(output_dir, recursive = TRUE)
output_file <- file.path(output_dir, "avg_ride_duration_by_day.csv")

# Calculate average ride duration by rider type and day, filtering ride_length <= 36.0
avg_ride_duration <- eda_data %>%
  filter(!is_outlier) %>%
  group_by(member_casual, day_of_week) %>%
  summarise(
    avg_duration = mean(ride_length, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  mutate(
    avg_duration = round(avg_duration, 1),
    day_of_week = factor(day_of_week, levels = c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"))
  )

# Display table
avg_ride_duration %>%
  pivot_wider(
    names_from = member_casual,
    values_from = avg_duration
  ) %>%
  knitr::kable(caption = "Average Ride Duration (Minutes) by Day and Rider Type") %>%
  kableExtra::kable_styling(full_width = FALSE, position = "left")

Average Ride Duration (Minutes) by Day and Rider Type
day_of_week	casual	member
Sunday	14.7	11.5
Monday	12.9	10.4
Tuesday	12.4	10.5
Wednesday	12.6	10.7
Thursday	12.6	10.5
Friday	13.4	10.5
Saturday	15.0	11.6

# Export to CSV
write_csv(avg_ride_duration, output_file)

Visualization The bar chart below compares the average ride duration by day for each rider type.

Results:
Casual riders consistently take longer trips than members, regardless of the day of the week. Weekdays show stable, shorter ride durations for both groups, reflecting typical commuting or utility trips. On weekends, both groups extend their rides somewhat, but the effect is much more pronounced among casual riders—suggesting more recreational or leisure-oriented usage patterns. Overall there is not a large enough difference in ride time to suggest action.

7.5 Time of day usage

To examine ride timing behavior, we analyzed when rides occur throughout the day, grouped into 2-hour intervals. This analysis includes only typical rides (≤ 36 minutes), as defined in Section 7.2, to avoid distortion from outlier activity. The dot chart below shows the proportion of rides per time block, by day of the week and rider type. Dot size indicates the share of rides for that day and rider group. Color intensity reflects each rider group’s daily peak activity — the darkest dots mark the most active time blocks.

Show Code

library(dplyr)
library(ggplot2)
library(scales)

# Define rider colors if not already set
cyclistics_colors <- c(
  "member" = "#1f77b4",  # Blue
  "casual" = "#ff7f0e"   # Orange
)

# Prepare and count hourly ride shares
hourly_data <- eda_data %>%
  filter(!is_outlier, !is.na(hour_of_day)) %>%
  mutate(
    hour_block = floor(hour_of_day / 2) * 2,
    time_label = sprintf("%02d–%02d %s",
                         ifelse(hour_block == 0, 12, ifelse(hour_block > 12, hour_block - 12, hour_block)),
                         ifelse(hour_block + 2 == 12, 12, ifelse(hour_block + 2 > 12, hour_block + 2 - 12, hour_block + 2)),
                         ifelse(hour_block < 12, "AM", "PM")),
    time_label = factor(time_label, levels = unique(time_label[order(hour_block)]))
  ) %>%
  count(member_casual, day_of_week, time_label) %>%
  group_by(member_casual, day_of_week) %>%
  mutate(
    share = n / sum(n),
    max_share = max(share),
    intensity = share / max_share,
    label = if_else(share == max_share, paste0(round(share * 100), "%"), NA_character_)
  ) %>%
  ungroup()

# Set factor order
hourly_data$day_of_week <- factor(hourly_data$day_of_week,
  levels = c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"))

# Apply group-specific intensity scaling using official rider colors
hourly_data$fill_color <- mapply(function(user, val) {
  col <- cyclistics_colors[[user]]
  base <- col2rgb(col) / 255
  rgb(1 - (1 - base[1]) * val,
    1 - (1 - base[2]) * val,
    1 - (1 - base[3]) * val)
}, hourly_data$member_casual, hourly_data$intensity)


# Build plot
dot_plot <- ggplot(hourly_data, aes(x = day_of_week, y = time_label)) +
      geom_tile(
    data = data.frame(
      member_casual = rep(c("member", "casual"), each = 2),
      day_of_week = rep(c("Saturday", "Sunday"), 2),
      time_label = rep(levels(hourly_data$time_label)[1], 4)
    ),
    aes(x = day_of_week, y = time_label),
    width = 1, height = Inf,
    fill = "darkgreen", alpha = 0.1,
    inherit.aes = FALSE
  ) +
  # Weekend label
  annotate("text", x = 6.5, y = tail(levels(hourly_data$time_label), .5),
           label = "Weekend", fontface = "plain", size = 3.5,
           color = "darkgreen", vjust = -1, clip = "off") +
  # Plot layers
  geom_point(aes(size = share, fill = fill_color), shape = 21,
           color = "black", stroke = 0.3, show.legend = FALSE) +
  # Outline layer
geom_text(aes(label = label), size = 4.5, color = "black", fontface = "bold", na.rm = TRUE) +
# Foreground layer
geom_text(aes(label = label), size = 4.5, color = "white", fontface = "bold", na.rm = TRUE) +
  
  scale_fill_identity() +
  scale_size_continuous(range = c(1.5, 13)) +
  facet_wrap(~ member_casual, ncol = 1) +
  labs(
    title = "Top Ride Time Blocks by Rider Type",
    subtitle = "Dot size and color show each time block’s share of daily rides by rider type \n — highlighting peak periods averaged across the year.",
    x = "Day of Week",
    y = "Time of Day (2-Hour Intervals)"
  ) +
  theme_minimal(base_size = 11) +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1),
  strip.text = element_text(face = "bold", size = 14),
  panel.spacing = unit(1, "lines"),
  plot.margin = margin(t = 30, r = 10, b = 10, l = 10),
  plot.title = element_text(face = "bold", hjust = 0.5),
  plot.subtitle = element_text(hjust = 0.5),
  clip = "off"
  )

# Save plot
ggsave("hourly_dot_chart_v9_web.png", plot = dot_plot, width = 8, height = 12, dpi = 300)

Results:
Member rides exhibit clear peaks during weekday morning commute hours (7–9 AM) and evening commute hours (4–6 PM), consistent with regular workday travel patterns. In contrast, casual riders show a broader peak in midday usage (11 AM to 3 PM) and heightened activity on weekends, reflecting more flexible and recreational ride behavior. This distinct usage pattern underscores the routine commuting nature of members versus the leisure-oriented preferences of casual riders.

7.6 Station Usage

This section identifies the most active stations in the Cyclistic network for both casual riders and annual members. By mapping and ranking the top 10 stations used by each group, we highlight geographic and behavioral differences in ridership. These insights support targeted marketing and operational decisions, ensuring that promotional efforts and resource allocation reflect real-world usage patterns.

Show Code

library(dplyr)
library(tidyr)
library(knitr)
library(kableExtra)
library(scales)

cyclistics_colors <- c(
  "member" = "#1f77b4",   # Blue
  "casual" = "#ff7f0e"    # Orange
)

# Add or refresh 'is_outlier' column for consistent filtering throughout analysis
if (!"is_outlier" %in% names(eda_data)) {
  eda_data <- eda_data %>%
    mutate(is_outlier = ride_length > 36.0)
}

# Table function
get_top_stations <- function(data, group_col, station_col) {
  out <- data %>%
    filter(member_casual == group_col) %>%
    count({{station_col}}, sort = TRUE) %>%
    slice_head(n = 10) %>%
    rename(Station = {{station_col}}, Rides = n)
  out$Rides <- comma(out$Rides)
  out
}

# Top 10 locations for map plotting (combines starts & ends)
get_top10_usage_locations <- function(data, group_col) {
  data %>%
    filter(member_casual == group_col) %>%
    select(start_station_name, start_lat, start_lng, end_station_name, end_lat, end_lng) %>%
    pivot_longer(
      cols = starts_with("start_station_name"):ends_with("end_lng"),
      names_to = c("type", ".value"),
      names_pattern = "(start|end)_(station_name|lat|lng)"
    ) %>%
    count(station_name, lat, lng, sort = TRUE) %>%
    slice_head(n = 10)
}

# Generate tables for display
casual_start <- get_top_stations(eda_data, "casual", start_station_name)
casual_end   <- get_top_stations(eda_data, "casual", end_station_name)
member_start <- get_top_stations(eda_data, "member", start_station_name)
member_end   <- get_top_stations(eda_data, "member", end_station_name)

# Overlap for columns
casual_start_names <- casual_start$Station
casual_end_names   <- casual_end$Station
member_start_names <- member_start$Station
member_end_names   <- member_end$Station

# Casual Table with Overlap
casual_tbl <- cbind(
  "Top 10 Start Stations" = casual_start$Station,
  "Rides (Start)" = casual_start$Rides,
  "Also Top 10 End?" = ifelse(casual_start$Station %in% casual_end_names, "✔", ""),
  "Top 10 End Stations" = casual_end$Station,
  "Rides (End)" = casual_end$Rides,
  "Also Top 10 Start?" = ifelse(casual_end$Station %in% casual_start_names, "✔", "")
)

# Member Table with Overlap
member_tbl <- cbind(
  "Top 10 Start Stations" = member_start$Station,
  "Rides (Start)" = member_start$Rides,
  "Also Top 10 End?" = ifelse(member_start$Station %in% member_end_names, "✔", ""),
  "Top 10 End Stations" = member_end$Station,
  "Rides (End)" = member_end$Rides,
  "Also Top 10 Start?" = ifelse(member_end$Station %in% member_start_names, "✔", "")
)

# Output tables
cat("**Casual Riders**\n")

## **Casual Riders**

kable(casual_tbl, align = "lrrlrr") %>% kable_styling(full_width = FALSE)

Top 10 Start Stations	Rides (Start)	Also Top 10 End?	Top 10 End Stations	Rides (End)	Also Top 10 Start?
Streeter Dr & Grand Ave	47,457	✔	Streeter Dr & Grand Ave	51,579	✔
DuSable Lake Shore Dr & Monroe St	31,573	✔	DuSable Lake Shore Dr & Monroe St	29,496	✔
Michigan Ave & Oak St	22,992	✔	DuSable Lake Shore Dr & North Blvd	24,857	✔
DuSable Lake Shore Dr & North Blvd	21,087	✔	Michigan Ave & Oak St	23,886	✔
Millennium Park	20,469	✔	Millennium Park	22,543	✔
Shedd Aquarium	19,779	✔	Shedd Aquarium	17,973	✔
Dusable Harbor	16,992	✔	Theater on the Lake	16,713	✔
Theater on the Lake	15,205	✔	Dusable Harbor	15,403	✔
Michigan Ave & 8th St	12,383	✔	Michigan Ave & 8th St	11,478	✔
Adler Planetarium	12,051		Michigan Ave & Washington St	10,568

cat("\n\n**Member Riders**\n")

## 
## 
## **Member Riders**

kable(member_tbl, align = "lrrlrr") %>% kable_styling(full_width = FALSE)

Top 10 Start Stations	Rides (Start)	Also Top 10 End?	Top 10 End Stations	Rides (End)	Also Top 10 Start?
Kingsbury St & Kinzie St	26,525	✔	Kingsbury St & Kinzie St	26,620	✔
Clinton St & Washington Blvd	24,619	✔	Clinton St & Washington Blvd	25,127	✔
Clark St & Elm St	22,106	✔	Clinton St & Madison St	23,063	✔
Clinton St & Madison St	22,013	✔	Clark St & Elm St	21,991	✔
Clinton St & Jackson Blvd	18,264	✔	Clinton St & Jackson Blvd	18,211	✔
Wells St & Concord Ln	17,966	✔	Wells St & Concord Ln	18,133	✔
Wells St & Elm St	17,728	✔	Wells St & Elm St	17,758	✔
Dearborn St & Erie St	17,370		University Ave & 57th St	17,577	✔
University Ave & 57th St	17,206	✔	State St & Chicago Ave	17,319
Canal St & Madison St	16,863	✔	Canal St & Madison St	16,621	✔

# Get map data for both groups
top10_member <- get_top10_usage_locations(eda_data, "member")
top10_casual <- get_top10_usage_locations(eda_data, "casual")

top10_casual$dot_id <- 1:nrow(top10_casual)
top10_member$dot_id <- 1:nrow(top10_member)

cyclistics_colors <- c(
  "member" = "#1f77b4",   # Blue
  "casual" = "#ff7f0e"    # Orange
)
# (rest of your code for map data prep...)


# Bounding box covering all
all_lng <- c(top10_member$lng, top10_casual$lng)
all_lat <- c(top10_member$lat, top10_casual$lat)
bbox <- c(
  left   = min(all_lng) - 0.03,
  bottom = min(all_lat) - 0.01,
  right  = max(all_lng) + 0.03,
  top    = max(all_lat) + 0.01
)

# Download map once for both
library(ggmap)
library(ggrepel)
register_stadiamaps(key = "0cde5034-d815-4778-a1f8-b1d15eb70b0b")
chicago_map <- get_stadiamap(
  bbox = bbox,
  zoom = 13,
  maptype = "outdoors",
  size = c(1500, 3000) # Tall/narrow map tiles
)

Dot	Station Name	Total Uses
1	Streeter Dr & Grand Ave	94,524
2	DuSable Lake Shore Dr & Monroe St	57,915
3	Michigan Ave & Oak St	44,374
4	DuSable Lake Shore Dr & North Blvd	43,984
5	Millennium Park	40,730
6	Shedd Aquarium	35,452
7	Dusable Harbor	30,511
8	Theater on the Lake	30,407
9	Michigan Ave & 8th St	22,660
10	Adler Planetarium	21,017

Dot	Station Name	Total Uses
1	Kingsbury St & Kinzie St	48,664
2	Clinton St & Washington Blvd	44,252
3	Clinton St & Madison St	40,726
4	Clark St & Elm St	40,624
5	University Ave & 57th St	33,303
6	Wells St & Concord Ln	33,083
7	Clinton St & Jackson Blvd	32,961
8	Wells St & Elm St	32,197
9	State St & Chicago Ave	31,127
10	Dearborn St & Erie St	31,013

Results:
The maps and tables above reveal that casual riders predominantly use stations near major tourist attractions and along the lakeshore, such as Streeter Dr & Grand Ave, Millennium Park, and Shedd Aquarium. Annual members favor stations in business districts and transit hubs, such as Kingsbury St & Kinzie St and Clinton St & Washington Blvd.

Both groups have overlapping top stations, but the spatial patterns and volume differences underscore the recreational focus of casual riders and the routine, commuter-oriented patterns of members. These findings support targeted marketing strategies by location and rider type.

7.7 Casual User Hotspots

This section identifies the top 10 stations most frequently used by casual riders on weekends. By analyzing hourly ride patterns at these hotspots, we can better understand when and where casual riders are most active, supporting more targeted marketing and outreach efforts.

Show Code

library(dplyr)
library(ggplot2)
library(lubridate)

# 1. Prepare data: filter to weekends, casual riders, top 10 locations
plot_data <- eda_data %>%
    filter(
        member_casual == "casual",
        day_of_week %in% c("Saturday", "Sunday"),
        !is.na(start_station_name)
    ) %>%
    mutate(hour_of_day = hour(started_at)) %>%
    group_by(start_station_name, hour_of_day) %>%
    summarise(rides = n(), .groups = "drop")

# 2. Identify top 10 locations by total rides
top_stations <- plot_data %>%
    group_by(start_station_name) %>%
    summarise(total_rides = sum(rides), .groups = "drop") %>%
    arrange(desc(total_rides)) %>%
    slice_head(n = 10) %>%
    pull(start_station_name)

plot_data_top <- plot_data %>%
    filter(start_station_name %in% top_stations) %>%
    mutate(
        start_station_name = factor(
            start_station_name,
            levels = top_stations  # Ensures ordered by total rides
        )
    )

# 3. Filter hours 8–22
plot_data_filtered <- plot_data_top %>% 
    filter(hour_of_day >= 8 & hour_of_day <= 22)

# 4. Generate and save plot
my_plot <- ggplot(plot_data_filtered, aes(x = hour_of_day, y = rides, color = start_station_name, group = start_station_name)) +
    geom_line(size = 1) +
    facet_wrap(~ start_station_name, ncol = 2, scales = "free_y") +
    labs(
        title = "Most Used Locations by Hour (Weekends, Casual Users)",
        subtitle = "Top 10 locations by total rides, Saturdays and Sundays only",
        x = "Hour of Day",
        y = "Number of Rides",
        color = "Location"
    ) +
    scale_x_continuous(
        breaks = 8:22,
        labels = as.character(8:22)
    ) +
    theme_minimal(base_size = 13) +
    theme(legend.position = "none")

ggsave("weekend_top_stations.png", plot = my_plot, width = 10, height = 8, dpi = 150)

Results:
The analysis shows that weekend activity among casual riders is highly concentrated at a handful of stations. These top 10 locations experience the highest ride volumes during core daytime hours, particularly from late morning through early evening. Peaks in ridership typically occur between 11 AM and 6 PM. Focusing marketing resources at these hotspots—especially during peak hours—will maximize visibility and engagement with casual users.

7.8 Bike type Usage

This section compares usage patterns for classic bikes, electric bikes, and scooters across rider types.

Show Code

library(dplyr)
library(scales)
library(ggplot2)
library(ggtext)

# Data prep
bar_data <- eda_data %>%
  filter(!is_outlier) %>%
  group_by(member_casual, rideable_type) %>%
  summarise(Rides = n(), .groups = "drop") %>%
  group_by(member_casual) %>%
  mutate(
    Share = Rides / sum(Rides),
    label = paste0("<b>", percent(Share, accuracy = 1), "</b><br>", comma(Rides)),
    label_y = Rides + max(Rides) * 0.04 # to show label in correct position
  ) %>%
  ungroup()

bike_type_labels <- c(
  classic_bike = "Classic\nBike",
  electric_bike = "Electric\nBike",
  electric_scooter = "Electric\nScooter"
)

cyclistics_colors <- c(
  "member" = "#1f77b4",   # Blue
  "casual" = "#ff7f0e"    # Orange
)

bike_type_palette <- c(
  "classic_bike" = "#8d5524",
  "electric_bike" = "#7B9E87",
  "electric_scooter" = "#BFA980"
)

bike_type_emoji <- c(
  classic_bike = "🚲",
  electric_bike = "⚡🚲",
  electric_scooter = "🛴"
)

bike_type_plot <- ggplot(
    bar_data,
    aes(x = member_casual, y = Rides, fill = rideable_type)
  ) +
  geom_col(
    position = position_dodge(width = 0.8), width = 0.7, show.legend = FALSE
  ) +
  ggtext::geom_richtext(
    aes(y = label_y, label = label, group = rideable_type),
    position = position_dodge(width = 0.8),
    vjust = 0,
    fill = NA, label.color = NA,
    color = "black", size = 4, lineheight = 1.1
  ) +
  geom_text(
    aes(
      label = bike_type_emoji[as.character(rideable_type)],
      y = 3000   
    ),
    position = position_dodge(width = 0.8),
    size = 7,
    vjust = 0,
    family = "Segoe UI Emoji" # Remove or change for Mac/Linux
  ) +
  scale_fill_manual(
    values = bike_type_palette
  ) +
  scale_x_discrete(
  labels = c(
    casual = "<span style='color:#ff7f0e; font-size:18px;'>&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;</span><br><br><b>Casual</b>",
    member = "<span style='color:#1f77b4; font-size:18px;'>&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;&#9608;</span><br><br><b>Member</b>"
  )
) +
  scale_y_continuous(
    labels = scales::comma,
    trans = "sqrt",
    expand = expansion(mult = c(0, 0.18))
  ) +
  labs(
    title = "Ride Counts by Rider and Bike Type",
    x = "Rider Type",
    y = "Number of Rides"
  ) +
  theme_minimal(base_size = 14) +
  theme(
    plot.title = element_text(
      hjust = 0.5, size = 18, face = "bold", margin = margin(b = 20)
    ),
    axis.title.x = element_text(size = 16, margin = margin(t = 12)),
    axis.text.x = ggtext::element_markdown(size = 16, face = "bold"),
    legend.position = "none"
  )

ggsave("bike_type_usage_v9.png", plot = bike_type_plot, width = 8, height = 5, dpi = 150)

Results:
Classic bikes (Brown bars) are by far the most popular vehicle choice for both members and casual riders, representing a clear preference across all users. Electric bikes (Green bars) are a strong secondary option, with notably higher adoption among casual riders than members. While electric scooters (Tan bars) are available, they make up only a tiny fraction of trips for both groups.

Key insights: Classic bikes dominate overall usage and are the primary mode for both segments. Casual riders show greater diversity in vehicle selection, opting for electric bikes and scooters more frequently than members. Electric scooters have limited appeal, indicating that most riders, regardless of membership status, favor bikes over scooters for their trips.

8 Key Findings

Casual riders primarily use the service near popular attractions and high-traffic areas, indicating strong demand from tourists and recreational users.

Casual riders consistently take longer and more varied trips than members, with the biggest differences on weekends—highlighting leisure-oriented usage.

Casual usage is more evenly distributed throughout the day and peaks on weekends and afternoons, in contrast to members, whose rides cluster around traditional commute hours.

Seasonal patterns are evident, with overall usage peaking during warmer months and at central stations that serve both visitors and commuters.

Members use the system more frequently and for shorter, regular trips, consistent with commuting and utilitarian travel.

Excluding outlier rides (>36 minutes) sharpened the behavioral comparison, focusing the analysis on typical ride patterns.

Bike type differences are notable: Classic bikes remain the top choice for all users, electric bikes are more popular among casual riders but still secondary, and scooters account for only a tiny share of total rides.

9. Recommendations

Enhance Station-Level Marketing Using Location Data

To further optimize marketing strategies and drive casual ridership and conversion to members we recommend leveraging the Google Places API (or similar external data sources) to identify restaurants, bars, attractions, and other points of interest within a half-mile radius of the most-used stations. This approach can help:

-Target promotional campaigns at locations with high tourist or dining activity.

-Inform partnership opportunities with local businesses.

-Support decisions on seasonal station rebalancing or placement.

This location-based enrichment can provide the marketing team with actionable intelligence for hyper-local outreach, sponsorships, and cross-promotion efforts.

(See Appendix for an example workflow and R code.)

10. Limitations

This analysis is limited to Cyclistics ride data from 2024. Individual rider account details and pricing information were not available, which restricted the ability to assess user-level behavior or conduct pricing sensitivity analysis. Access to these additional data sources would enable a deeper understanding of rider segments, trip motivations, and price responsiveness.

11. Appendix

This workflow is an example for the marketing team identifying high-value partnership opportunities and craft targeted outreach based on actual neighborhood business activity.

Table: Example of restaurants, bars, and attractions within 0.5 miles of top casual rider stations (Google Places API, May 2025)

Show Code

library(dplyr)
library(readr)

top10_casual_with_locs <- eda_data %>%
  filter(member_casual == "casual", !is.na(start_station_name)) %>%
  count(start_station_name, sort = TRUE) %>%
  slice_head(n = 10) %>%
  left_join(
    eda_data %>%
      group_by(start_station_name) %>%
      summarize(
        start_lat = first(na.omit(start_lat)),
        start_lng = first(na.omit(start_lng)),
        .groups = "drop"
      ),
    by = "start_station_name"
  ) %>%
  mutate(
    start_lat = sprintf("%.5f", as.numeric(start_lat)),
    start_lng = sprintf("%.5f", as.numeric(start_lng))
  )

write_csv(top10_casual_with_locs, "appendix_top10_casual_2024.csv")

library(httr)
library(jsonlite)
library(readr)
library(dplyr)
library(purrr)

api_key <- Sys.getenv("GOOGLE_API_KEY") # Use environmental variable
search_radius <- 800  # About 0.5 miles in meters

top10 <- read_csv("appendix_top10_casual_2024.csv")

get_places <- function(lat, lng) {
  url <- paste0(
    "https://maps.googleapis.com/maps/api/place/nearbysearch/json?",
    "location=", lat, ",", lng,
    "&radius=", search_radius,
    "&type=restaurant|bar|tourist_attraction",
    "&key=", api_key
  )
  response <- httr::GET(url)
  content <- httr::content(response, as = "text", encoding = "UTF-8")
  json <- jsonlite::fromJSON(content, flatten = TRUE)
  results <- json$results
  if (is.null(results) || nrow(results) == 0) return(NULL)
  
  n <- nrow(results)
  # Defensive extraction: handles missing, empty, or weird structures
  tibble(
    name = if (!is.null(results$name)) results$name else rep(NA_character_, n),
    address = if (!is.null(results$vicinity)) results$vicinity else rep(NA_character_, n),
    place_type = if (!is.null(results$types)) sapply(results$types, function(x) if (length(x) == 0) NA_character_ else paste(x, collapse = ", ")) else rep(NA_character_, n),
    lat = if (!is.null(results$geometry.location.lat)) results$geometry.location.lat else rep(NA_real_, n),
    lng = if (!is.null(results$geometry.location.lng)) results$geometry.location.lng else rep(NA_real_, n)
  )
}

# Loop over each top station, collecting places
all_places <- purrr::pmap_dfr(
  top10[, c("start_station_name", "start_lat", "start_lng")],
  function(start_station_name, start_lat, start_lng) {
    places <- get_places(start_lat, start_lng)
    if (!is.null(places)) {
      places$station <- start_station_name
      return(places)
    } else {
      return(NULL)
    }
  }
)

write_csv(all_places, "appendix_places_nearby_top10.csv")

library(readr)
library(knitr)
library(kableExtra)

if (file.exists("appendix_places_nearby_top10.csv")) {
  all_places <- read_csv("appendix_places_nearby_top10.csv", show_col_types = FALSE)
  if (nrow(all_places) > 0) {
    # Select & rename columns for readability if desired
    colnames(all_places) <- gsub("_", " ", colnames(all_places))
    kable(
      all_places,
      caption = "Restaurants, Bars, and Attractions Near Top 10 Casual Rider Stations (Google Places API, May 2025)",
      align = "l"
    ) %>%
    kable_styling(
      bootstrap_options = c("striped", "hover", "condensed"),
      full_width = FALSE,
      position = "left"
    ) %>%
    scroll_box(height = "350px")
  } else {
    cat("No nearby places found in the CSV. Please check your enrichment results.")
  }
} else {
  cat("API enrichment table not yet generated. Run the save chunk first.")
}

Cyclistics Capstone Analysis

Jonathan Donovan

2025-05-05

1. Executive Summary

2. About Cyclistic

3. Key Roles

4. Business Task

Specific Questions Answered

5. Data Source, Tools and Setup

5.1 Data Source and Credibility

5.2 Tools and Packages

5.3 Load Raw Dataset

5.4 Create Working Dataset

6. Data Cleaning

6.1 Identify Rows with Insufficient Data

6.2 Verify Data Types

6.3 Verify Categorical Fields

6.4 Validate Minimum Required Fields

6.5 Count Records to be Removed

6.6 Check for Duplicate Ride IDs

6.7 Identify Short Rides

6.8 Apply Validation Filters and generate data for exploratory analysis.

7. Analysis and Findings

7.1 Initial Look - Outliers

7.2 Monthly Ride Volume by rider type

7.3 Total Rides by Rider Type

7.4 Average Ride Duration by Day

7.5 Time of day usage

7.6 Station Usage

7.7 Casual User Hotspots

7.8 Bike type Usage

8 Key Findings

9. Recommendations

10. Limitations

11. Appendix